Teaching design generation method, electronic device, and readable storage medium

Abstract

The application belongs to the technical field of digital education, and discloses a teaching design generation method, an electronic device and a readable storage medium. The method comprises the following steps: determining a target prompt; the target prompt carries at least a teaching element; based on the target prompt, a large language model is called to generate a teaching design outline corresponding to the teaching element; from a second intelligent agent deployed on a second device, a third intelligent agent corresponding to each teaching element is called respectively to generate an optimized text corresponding to an initial text corresponding to each teaching element in the teaching design outline; each second intelligent agent can search in a teaching design related knowledge base corresponding to the teaching element; and the teaching design outline and all the optimized texts are integrated to obtain a target teaching design. The method improves the usability of the teaching design generated by the large language model.

Description

Technical Field

[0001] This application relates to the field of digital education technology, specifically to an instructional design generation method, electronic device, and readable storage medium. Background Technology

[0002] With the rapid development of information technology, the education sector is undergoing a profound digital transformation. This transformation is primarily reflected in the digitization of teaching content, the online transformation of teaching methods, and the intelligentization of teaching management. Against this backdrop, instructional design, as a crucial element in the teaching process, is increasingly incorporating innovation and application of intelligent technologies. For example, in instructional design, LLM (Large Language Model) can assist teachers in quickly generating teaching plans.

[0003] However, current instructional designs generated by large language models typically only list knowledge points, making it difficult for teachers to use these designs for instruction. In other words, the usability of instructional designs generated by large language models is currently poor. Summary of the Invention

[0004] The purpose of this application is to provide a method for generating instructional designs, an electronic device, and a readable storage medium, aiming to solve the problem of how to improve the usability of instructional designs generated from large language models.

[0005] To solve the above-mentioned technical problems, this application is implemented as follows:

[0006] In a first aspect, embodiments of this application provide a method for generating instructional designs, applied to a first device on which a first intelligent agent is deployed, the method comprising:

[0007] Define target prompts; the target prompts should at least contain teaching elements;

[0008] Based on the target prompts, the large language model is invoked to generate the teaching design outlines corresponding to the teaching elements;

[0009] From the second agent deployed on the second device, the third agent corresponding to each teaching element is invoked to generate the optimized text corresponding to the initial text of each teaching element in the teaching design outline; each second agent can search in the teaching design-related knowledge base corresponding to the teaching element.

[0010] The aforementioned instructional design outline and all optimized texts are integrated to obtain the target instructional design.

[0011] Optionally, the determination of the target prompt includes:

[0012] In response to the user's input of a requirement description, an initial prompt corresponding to the requirement description is generated;

[0013] The target prompt is determined based on the initial prompt.

[0014] Optionally, determining the target prompt based on the initial prompt includes:

[0015] Determine a first number of teaching elements in the requirements description, and determine a second number of teaching elements in the initial prompt;

[0016] The second quantity is scored based on the first quantity to obtain a first sub-score;

[0017] If the first sub-score is greater than or equal to the first score threshold, then the generation order of each teaching element in the initial prompt and the corresponding prompt text are determined.

[0018] The generation order is scored to obtain a second sub-score;

[0019] The content consistency of the prompt texts corresponding to each adjacent teaching element is scored separately to obtain the third sub-score;

[0020] If the sum of the second sub-score and the third sub-score is greater than or equal to the second score threshold, then the initial prompt is optimized to obtain the target prompt.

[0021] Optionally, the step of scoring the second quantity based on the first quantity to obtain a first sub-score further includes:

[0022] If the first sub-score is less than the first score threshold, then return to the initial prompt step corresponding to generating the requirement description.

[0023] Optionally, the step of scoring the content consistency of the prompt texts corresponding to each adjacent teaching element to obtain a third sub-score further includes:

[0024] If the sum of the second sub-score and the third sub-score is less than the second score threshold, then return to the initial prompt step corresponding to generating the requirement description.

[0025] Optionally, optimizing the initial prompt to obtain the target prompt includes:

[0026] Determine the user's requirement for secondary input;

[0027] Based on the requirements described by the secondary input, the large language model is invoked to optimize the initial prompt, thereby obtaining the target prompt.

[0028] Secondly, embodiments of this application also provide a method for generating instructional designs, applied to a second device deployed with a second intelligent agent, and for any third intelligent agent among the second intelligent agents, the method includes:

[0029] In response to a call instruction sent by a first intelligent agent deployed on a first device, the initial text corresponding to the target teaching element in the instructional design syllabus carried in the call instruction is determined; the target teaching element corresponds to any of the third intelligent agents.

[0030] The initial text is optimized to generate optimized text;

[0031] The optimized text is sent to the first intelligent agent; the optimized text is used to integrate with the instructional design outline to obtain the target instructional design.

[0032] Optionally, the step of optimizing the initial text to obtain optimized text includes:

[0033] Based on the target teaching elements and the initial text, a search is performed in the instructional design-related knowledge base to obtain the recall text;

[0034] The initial text is optimized based on the recalled text to obtain the intermediate text;

[0035] If the intermediate text meets the optimization requirements, the intermediate text is determined to be the optimized text.

[0036] Thirdly, embodiments of this application provide an electronic device, which includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the steps of the method described in the first aspect.

[0037] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0038] Compared to related technologies, this application's embodiments generate instructional design outlines corresponding to educational elements through a large language model. Then, from a second agent deployed on a second device, a third agent corresponding to each instructional element is invoked to generate optimized text corresponding to the initial text of each instructional element in the instructional design outline. Each second agent can then search within the instructional design-related knowledge base corresponding to the instructional element. In other words, using instructional elements as the basic unit, and based on the third agent's search within the corresponding instructional design-related knowledge base, optimized text corresponding to the initial text is generated, resulting in in-depth optimized text. For example, if the instructional element is instructional reflection design, after optimizing the text through the corresponding instructional design-related knowledge base, an in-depth instructional design related to instructional reflection design can be obtained. Subsequently, by integrating the instructional design outline and all optimized texts, the target instructional design is obtained. This means that the content corresponding to any instructional element in the target instructional design (such as student learning analysis, teaching objectives, etc.) has been optimized using knowledge from the relevant instructional design-related knowledge base. This makes the instructional design include not only knowledge points but also a deeper level of how to teach, thereby improving the usability of the instructional design generated through the large language model. Attached Figure Description

[0039] Figure 1 A flowchart illustrating the instructional design generation method provided in this application embodiment;

[0040] Figure 2 Another flowchart illustrating the instructional design generation method provided in this application embodiment;

[0041] Figure 3 This is a schematic diagram of the instructional design generation system provided in an embodiment of this application. Detailed Implementation

[0042] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0043] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0044] To illustrate the technical solution described in this application, specific embodiments are provided below.

[0045] As mentioned in the background section, current instructional designs for large language model generation typically only list knowledge points, making it difficult for teachers to use them effectively. The reason for this is that these designs inherently lack depth; they only present knowledge points without addressing how to teach them, such as the design of interactive and reflective activities. Therefore, they fail to effectively stimulate student thinking and participation, resulting in poor usability of instructional designs for large language model generation.

[0046] To address the aforementioned issues, an exemplary embodiment provides a method for generating instructional designs. This method generates instructional design outlines corresponding to educational elements using a large language model. Then, from a second agent deployed on a second device, a third agent corresponding to each instructional element is invoked to generate optimized text corresponding to the initial text of each instructional element in the instructional design outline. Each second agent can retrieve relevant knowledge bases related to instructional designs for each instructional element. In other words, using instructional elements as the basic unit, optimized text corresponding to the initial text is generated based on the third agent's retrieval in the relevant instructional design knowledge base, resulting in in-depth optimized text. For example, if the instructional element is instructional reflection design, after optimizing the text through the relevant instructional design knowledge base, an in-depth instructional design related to instructional reflection design can be obtained. Subsequently, by integrating the instructional design outline and all optimized texts, a target instructional design is obtained. This means that the content corresponding to any instructional element in the target instructional design (such as student learning analysis, instructional objectives, etc.) has been optimized using knowledge from the relevant instructional design knowledge base. This allows instructional design to include not only knowledge points but also deeper aspects of how to teach, thereby improving the usability of instructional designs generated through large language models.

[0047] In this embodiment, the method is applied to a first device on which a first intelligent agent is deployed. This first intelligent agent can be referred to as a managerial intelligent agent, capable of managing the entire process of generating instructional designs. The first device can be a terminal, such as a computer or mobile phone; alternatively, it can be a server.

[0048] The manager agent can be constructed through the following steps S11 to S15.

[0049] Step S11: Obtain instructional design data and extract instructional elements.

[0050] First, data collection and preprocessing are performed.

[0051] I. Obtaining Instructional Design Data: Collect a large amount of raw instructional design data from multiple educational platforms (such as online course platforms), teaching resource databases (such as the National Public Service Platform for Educational Resources), and teacher feedback systems (such as evaluation data in the teaching management system) through methods such as web scraping, API interfaces, or database queries. This data should cover instructional design cases from different subjects (such as Chinese, mathematics, and Python programming), different grade levels (such as primary, middle, and high school), and different lesson types (such as new lessons and review lessons), as well as relevant supplementary information such as textbook content, curriculum standards, and teaching objectives. Ensure the diversity and representativeness of the data.

[0052] II. Data Cleaning: Cleaning the collected raw data to remove irrelevant information (such as advertisements and non-teaching content), process duplicate data (such as multiple versions of the same lesson plan), and correct obvious errors (such as formatting errors and garbled characters). For example, using rules or machine learning models to identify and delete text paragraphs unrelated to the instructional design, and deduplicating duplicate lesson plan records. Using data imputation techniques (such as mean, median, or model prediction) to handle missing values, or correcting outliers based on contextual information.

[0053] III. Data Formatting: Convert the cleaned teaching data into a unified structured or semi-structured format (such as JSON, XML, or a specific database schema). This includes standardizing field names (e.g., unifying "Teaching Objectives" and "Course Goals" into "Teaching Objectives"), standardizing data types (e.g., standardizing time data), defining clear field rules, and standardizing units of measurement and symbols. For example, organize all instructional design texts into structured data containing standard fields such as "Course Title," "Teaching Objectives," "Student Learning Analysis," "Teaching Process," and "Teaching Evaluation and Reflection."

[0054] Then, statistics on teaching elements are collected.

[0055] Based on the preprocessed structured instructional design data, natural language processing (NLP) techniques (such as keyword extraction, named entity recognition, and text classification) and content analysis methods were used to statistically analyze the core elements that constitute the instructional design, their frequency of occurrence, and typical expression methods.

[0056] Specifically, a large number of instructional design texts were analyzed to identify and summarize the general elements of the instructional design framework (i.e., instructional elements), which mainly include: instructional objectives (covering three dimensions: knowledge and skills, processes and methods, and affective attitudes and values); instructional content (including analysis of instructional focus and difficulties); student learning analysis (student background, foundation, interests, etc.); instructional process (detailed to each stage, such as course introduction, explanation of new knowledge, consolidation exercises, class summary, etc., and including teacher activities, student activities, design intentions, and time allocation); and instructional assessment and reflection (including formative assessment, summative assessment methods, and key points for reflection).

[0057] Statistical analysis revealed that these core elements are essential components of a complete and scientific instructional design, laying the foundation for the subsequent construction of a needs description and prompting framework for the manager's intelligent agent. For example, statistics showed that over 95% of high-quality lesson plans explicitly included three-dimensional learning objectives and detailed instructional process design.

[0058] Step S12: Based on the teaching elements, perform a first requirement description for the instructional design.

[0059] The goal of this step is to generate a clear and accurate requirement description to guide the managerial agent in understanding its task—generating a teaching design syllabus.

[0060] I. Define the theoretical foundation of instructional design: Specify the core educational theories that instructional design should follow, such as constructivism (emphasizing students' active construction of knowledge), Gagné's nine events of instruction (systematic design of the instructional process), and formative assessment theory (emphasizing process feedback and adjustment). These theories provide guiding principles for the structure and content of instructional design.

[0061] II. Utilizing Theoretical Guidance to Organize Teaching Elements Logically: Combine the extracted teaching elements (teaching objectives, teaching content, student learning analysis, teaching process, teaching assessment and reflection) with the selected theoretical foundation, clarifying the content requirements of each element and the logical relationships between them. For example, based on Gagné's theory, the teaching process should include stages such as attracting attention (corresponding to course introduction), providing learning guidance (corresponding to knowledge explanation), eliciting behavior (corresponding to consolidation exercises), and providing feedback (corresponding to teaching assessment). Teaching objectives should conform to constructivist principles, emphasizing the cultivation of students' abilities rather than the rote memorization of knowledge.

[0062] III. Generating the First Requirements Description Text: The above-mentioned instructional design framework requirements, including theoretical foundations, constituent elements, and organizational logic, are transformed into a natural language text—the first requirements description—using the capabilities of a large language model. This description will be provided as input to the agent creation platform. For example, a requirements description could be: "Please play the role of an instructional design expert proficient in constructivism and Gagné's theory of instructional events. Your task is to generate a well-structured instructional design outline based on the knowledge points provided by the user. This outline must include clear three-dimensional learning objectives, an analysis based on student learning, a teaching process consistent with Gagné's 9 events (including introduction, explanation, practice, and summary stages, and specifying teacher and student activities and design intentions), formative assessment points, and directions for instructional reflection. Please ensure that each part is logically clear and interconnected."

[0063] Step S13: Generate a prompt framework based on the first requirement description.

[0064] Using a low-code platform or agent creation platform that supports multi-agent development, the initial requirement description text generated in the previous step is input into the platform. Based on its built-in large language model and prompting engineering capabilities, the platform automatically or semi-automatically parses the requirement description and constructs it into a structured prompt framework. This framework typically employs a modular design, for example:

[0065] I. @Priming module: Initialization instructions, setting the behavioral tone of the agent. For example: "You will strictly follow the following instructions to perform the task and generate a teaching design outline."

[0066] II. The @Persona module: Defines the agent role. For example: "You are an experienced Python programming instructional design expert who is familiar with the cognitive characteristics of eighth-grade students."

[0067] III. @Audience module: Define the target audience. For example: "The generated instructional design is primarily aimed at eighth-grade students and their teachers."

[0068] IV. @ContextControl module: Set global rules and constraints. For example: "@Rules: Instructional design titles should be concise and engaging, avoiding overly technical terminology. Learning objectives must cover knowledge, skills, and affective dimensions. The teaching process must include teacher activities, student activities, and design intent."

[0069] The V.@Instruction module contains detailed task instructions, typically including multiple sub-modules corresponding to different teaching elements. Each sub-module contains:

[0070] InputVariable: Defines the input (such as knowledge points provided by the user).

[0071] Commands: Specific operational steps (such as analyzing knowledge points, setting teaching objectives, and designing teaching activities).

[0072] Rules: Rules for specific modules (e.g., learning objectives must conform to the SMART principle, and activity design must consider student interaction).

[0073] OutputVariable: Defines the output format (e.g., generating Markdown formatted text).

[0074] For example, the @Instruction module might contain instructions such as: "Based on the input knowledge point 'Basic Operations of Python Lists' and the target audience, formulate knowledge and skills objectives (mastering list creation, access, modification, and deletion operations), process and method objectives (understanding list applications through practical examples), and affective attitude and value objectives (cultivating logical thinking and problem-solving abilities)." It also sets the rule: "Objectives must be specific, measurable, and suitable for eighth-grade students."

[0075] Step S14: Optimize the prompt framework to ensure that the teaching elements meet the teaching requirements.

[0076] The initially generated prompting framework is tested and iteratively optimized to ensure that it can guide the agent to generate a high-quality, pedagogical design syllabus that conforms to teaching standards.

[0077] I. Functional Testing: Create a temporary intelligent agent using the unoptimized prompting framework, input different teaching knowledge points (such as "linear equation in one variable" and "principle of photosynthesis"), and generate several teaching design outline samples.

[0078] II. Evaluation and Analysis: The generated examples are evaluated by educational experts or using pre-defined instructional design evaluation criteria (such as coverage-based statistical algorithms to check whether the target dimensions and process steps are complete), identifying problems such as missing elements, unclear logic, and insufficient depth. For example, the evaluation may find that the generated learning objectives generally lack measurability, or that the teaching process lacks student activity design.

[0079] III. Prompt Optimization: Adjust the prompt framework based on the evaluation results.

[0080] Personalized adjustments: If the role description of @Persona is not precise enough, resulting in inconsistent generated content style, then refine the role definition (e.g., "Expert in Inquiry-Based Teaching Design for Middle School Mathematics"). If the definition of @Audience is broad, then specify it (e.g., "First-year middle school student, intermediate mathematics foundation"). Adjust the instructions in @Instruction to better suit the requirements of specific subjects and grade levels.

[0081] Adding missing elements: If the assessment finds that key elements (such as student learning analysis and teaching reflection) are missing, then the inclusion of these elements should be mandated in @ContextControl, and the corresponding generation instruction module should be added to @Instruction. For example, add the rule in @ContextControl: "Must include 'Student Learning Analysis' and 'Teaching Reflection' sections."

[0082] Improve content quality: If certain elements are of low quality (e.g., learning objectives do not conform to the SMART principle, or activity designs lack operability), add more specific instructions and rules to the corresponding @Instruction module. For example, add the following to the Rules for setting learning objectives: "The verbs for learning objectives should be clear and point to observable student behaviors; it is recommended to refer to Bloom's Taxonomy." Add the following to the Commands for designing the instructional process: "Design at least one group collaborative inquiry activity."

[0083] Step S15: Create the manager agent based on the optimized prompt framework.

[0084] Once the prompting framework has been optimized and can reliably generate a teaching design syllabus that meets the requirements, the manager agent is formally created.

[0085] I. Parameter settings: On the agent creation platform, select or configure the underlying large language model (such as DeepSeek-R1, a model optimized for the education field), and set appropriate parameters, such as temperature (to control the randomness of generation, it is recommended to set it to around 0.5 to balance creativity and stability), maximum output length (max tokens, such as 2048, to ensure that it can accommodate the complete outline), etc.

[0086] II. Publish the agent: Publish or save the optimized prompt framework and set parameters on the platform to complete the creation of the manager agent.

[0087] The above completes the explanation of creating the manager intelligent agent.

[0088] In this embodiment, as Figure 1 As shown, the method includes steps 102 to 108. Wherein:

[0089] Step 102: Determine the target prompt; the target prompt shall at least contain teaching elements.

[0090] It should be noted that when calling the large language model, the prompt can guide the large language model to generate corresponding content, and the target prompt is determined according to the prompt framework mentioned above.

[0091] The target prompt can be determined based on the user-input requirement description or through other means, such as automatically generated requirement descriptions. This embodiment does not limit how the target prompt is determined.

[0092] For example, determining a target prompt based on a user-inputted demand description specifically includes the following steps: in response to the user-inputted demand description, generating an initial prompt corresponding to the demand description; and determining a target prompt based on the initial prompt.

[0093] The requirement description can be configured using a specialized requirement parsing hint template to guide a large language model in extracting key information. This hint engineering allows the large language model (such as DeepSeek-R1) to semantically parse the requirement description and output structured requirement data, which is then used to populate the hint framework, generating the initial hint. Taking the structured output of requirement data in JSON format as an example, the specific parameters for the requirement data output are as follows:

[0094] {

[0095] "Priming": "Generate a complete instructional design syllabus",

[0096] "Persona": "Teacher",

[0097] "Audience": {"Student"},

[0098] "Instruction": ["Student Learning Analysis", "Teaching Objectives", "Teaching Methods", "Teaching Process", "Teaching Reflection"]

[0099] }

[0100] It should be noted that when generating the initial prompts corresponding to the requirement description, the text processing capabilities of large language models (such as DeepSeek-R1) can be utilized to generate prompt templates and fill the parameters of the requirement data into the corresponding positions of the predefined prompts (i.e., prompt frames).

[0101] For example, filling "generate a complete instructional design syllabus" into the Priming module of the prompt framework is used to stimulate the educational expertise of the large language model; filling "teacher" into the Persona module of the prompt framework is used to define the teacher role and professional competence of the administrator agent; filling {"student"} into the Audience module of the prompt framework is used to clarify the characteristics and needs of the target student group. Filling ["learning analysis", "teaching objectives", "teaching methods", "teaching process", "teaching reflection"] into the Instruction module of the prompt framework is used to determine detailed task instructions and output requirements.

[0102] The Instruction module employs a dynamic generation mechanism, automatically constructing subtasks based on the array of teaching elements in the requirement data.

[0103] Additionally, when filling in the Priming, Persona, and Audience modules in the generated prompt template, the parameter values ​​entered by the user twice must be retained. For example, "Priming" is expanded to: "You are an experienced subject teacher and need to generate a complete lesson plan syllabus based on ${user input}$.", where "${user input}$" is a placeholder. The "Instruction" module generates different sub-modules based on the teaching elements, and sets the first sub-module to receive user input and output output1, the second sub-module to receive output1, output2, and so on, until all elements in the teaching element array have been received.

[0104] The above describes the process of generating the initial prompt.

[0105] For determining the target prompt based on the initial prompt, this involves scoring the initial prompt for completeness and coherence. After the initial prompt passes both the completeness and coherence scores, it is optimized to obtain the target prompt. It's important to note that the completeness and coherence scores are scored in a specific order: completeness is scored first, followed by coherence. If the completeness score fails, coherence is not scored, and the initial prompt is regenerated. Only after the completeness score passes is the coherence score scored, and only after passing the coherence score is the initial prompt optimized to obtain the target prompt; otherwise, the initial prompt is regenerated.

[0106] The completeness score refers to whether the number of sub-modules in the "Instruction" is the same as the number of teaching elements in the requirements description. If they are the same, the completeness score is passed; if they are different, the completeness score is failed. Specifically, a first number of teaching elements in the requirements description and a second number of teaching elements in the initial prompt are determined. The second number is scored based on the first number to obtain a first sub-score. If the first sub-score is greater than or equal to a first score threshold, the completeness score is passed, and the process proceeds to the coherence score. If the first sub-score is less than the first score threshold, the initial prompt is regenerated.

[0107] The coherence scoring includes a scoring of the generation order of the aforementioned sub-modules and a scoring of the consistency of the content in the initial prompt. Specifically, the coherence scoring process is as follows: determine the generation order of each teaching element in the initial prompt and the corresponding prompt text; score the generation order to obtain a second sub-score; score the content consistency of the prompt text corresponding to each adjacent teaching element to obtain a third sub-score; if the sum of the second sub-score and the third sub-score is greater than or equal to a second score threshold, then optimize the initial prompt to obtain the target prompt; if the sum of the second sub-score and the third sub-score is less than the second score threshold, then regenerate the initial prompt.

[0108] It is understandable that the completeness score ensures the completeness of the content in the subsequent target instructional design generated based on the target prompts, and the coherence score ensures the coherence of the content.

[0109] For example, optimizing the initial prompt to obtain the target prompt means determining the user's secondary input requirement description, and based on the secondary input requirement description, calling the large language model to optimize the initial prompt to obtain the target prompt.

[0110] The user's secondary input description of needs can be grade level + subject + knowledge point. By leveraging the text optimization capabilities of the large language model, the initial prompts can be filled in based on the user's secondary input description of needs, thereby ensuring that each teaching element meets educational standards, subject requirements, and students' individual needs.

[0111] For example, the secondary input requirement description is "High School + Mathematics + Function Monotonicity". For the initial prompt "Priming", the original "You are an experienced subject teacher and need to generate a complete teaching design syllabus based on ${user input}$" is filled into "You are an experienced subject teacher and need to generate a complete teaching design syllabus based on the monotonicity of high school mathematical functions." In other words, the secondary input requirement description is used to fill the placeholder position.

[0112] Step 104: Based on the target prompts, invoke the large language model to generate the instructional design outlines corresponding to the teaching elements.

[0113] The large language model is invoked, and the specific content corresponding to each teaching element is generated sequentially from beginning to end according to the order of the sub-modules in the target prompt, thereby generating the teaching design outline. Specifically, the first sub-module in the Instruction module receives the user's input of grade + subject + knowledge point, generates the specific content of the corresponding teaching element, outputs output1, and stores it; the second sub-module in the Instruction module receives output1, generates the specific content of the corresponding teaching element based on it, outputs output2, and stores it; and so on, until the teaching design outline is fully constructed.

[0114] Step 106: From the second agent deployed on the second device, the third agent corresponding to each teaching element is invoked to generate the optimized text corresponding to the initial text of each teaching element in the teaching design outline; each second agent can search in the teaching design-related knowledge base corresponding to the teaching element.

[0115] The second intelligent agent is a dedicated intelligent agent, corresponding one-to-one with each teaching element. For example, teaching elements include learning analysis and teaching methods; learning analysis corresponds to dedicated intelligent agent 1, and teaching methods correspond to dedicated intelligent agent 2. Based on RAG (Retrieval-Augmented Generation) technology, the dedicated intelligent agents can perform searches in instructional design-related knowledge bases (e.g., if the teaching element is learning analysis, then the instructional design-related knowledge base is the learning analysis knowledge base, which stores data in JSON format) to obtain retrieved text, which is then used to optimize the instructional design syllabus.

[0116] It is understandable that when a third-party intelligent agent is invoked to optimize the initial text corresponding to each teaching element in the instructional design outline and generate the corresponding optimized text, different specialized intelligent agents search the corresponding knowledge base through the initial text in the instructional design outline to obtain the corresponding recall text.

[0117] For example, in the process of generating the optimized text corresponding to the initial text of each teaching element in the instructional design outline, the recalled text may be intermediate text, that is, not the final optimized text.

[0118] In this embodiment, the third intelligent agent can call the large language model to perform deep scoring on the intermediate text. If the score is greater than or equal to the third score threshold, that is, the intermediate text meets the optimization requirements, then the intermediate text is determined as optimized text; if the score is less than the third score threshold, that is, the intermediate text does not meet the optimization requirements, then the third intelligent agent will search again in the instructional design-related knowledge base corresponding to the teaching elements until the intermediate text meets the optimization requirements.

[0119] The second device can be a terminal, such as a computer or mobile phone; it can also be a server. It should be noted that the first and second devices can be the same device or different devices.

[0120] It should be noted that the managerial agent can intelligently identify teaching elements in the instructional design syllabus based on the reasoning ability of the large language model, and based on the API (Application Programming Interface), it can transmit semantically complete instructions (such as calling instructions) through the teaching element identification prompts and schedule the corresponding dedicated agents.

[0121] Specifically, for the identification of teaching elements, the administrator agent can use a large language model to analyze the content of the teaching design syllabus and identify the teaching elements that need optimization, based on the prompts for identification. Then, based on the generation order of the teaching elements, it calls the corresponding dedicated agents to optimize the order of the initial text. For example, the generation order could be: student learning analysis → teaching objectives → teaching methods → teaching process → teaching reflection.

[0122] Step 108: Integrate the instructional design outline and all optimized texts to obtain the target instructional design.

[0123] After the initial text corresponding to all teaching elements in the instructional design outline is optimized, the administrator agent can integrate the prompt templates, call and guide the large language model to reorganize the content of the instructional design outline and all optimized texts according to the standard instructional design format, thereby forming a complete instructional design.

[0124] In one exemplary embodiment, such as Figure 2 As shown, a method for generating instructional designs is also provided, applied to a second device with a second intelligent agent deployed thereon. For any third intelligent agent among the second intelligent agents, the method for generating instructional designs includes:

[0125] Step 202: In response to a call instruction sent by a first intelligent agent deployed on the first device, determine the initial text corresponding to the target teaching element in the instructional design syllabus carried in the call instruction; the target teaching element corresponds to any of the third intelligent agents;

[0126] Step 204: Optimize the initial text to obtain optimized text;

[0127] Step 206: Send the optimized text to the first intelligent agent; the optimized text is used to integrate with the instructional design outline to obtain the target instructional design.

[0128] For example, the step of optimizing the initial text to obtain optimized text includes:

[0129] Based on the target teaching elements and the initial text, a search is performed in the instructional design-related knowledge base to obtain the recall text;

[0130] The initial text is optimized based on the recalled text to obtain the intermediate text;

[0131] If the intermediate text meets the optimization requirements, the intermediate text is determined to be the optimized text.

[0132] In this embodiment, the interaction processes between the third agent and the first agent are basically the same as those described in the previous embodiments. To avoid repetition, they will not be repeated here.

[0133] The process of constructing a second intelligent agent includes the following steps:

[0134] Step S21: Based on the teaching design syllabus, design multiple optimization elements and clarify the role setting of the dedicated intelligent agent.

[0135] Based on the instructional design syllabus and common instructional design evaluation dimensions, key instructional elements requiring in-depth optimization are identified. A dedicated agent role is designed for each instructional element requiring optimization. These dedicated agents include, but are not limited to:

[0136] I. Learning Analysis Agent: Responsible for in-depth analysis of students' backgrounds, existing knowledge, learning styles, interests, etc., and generating personalized learning analysis reports.

[0137] II. Learning Objectives Refinement Agent: Responsible for refining the initial objectives in the syllabus into specific, measurable, achievable, relevant, and time-bound (SMART) learning objectives, covering the three dimensions of knowledge, skills, and affect, and analyzing the key points and difficulties in teaching.

[0138] III. Teaching Process Generating Agent: Responsible for selecting appropriate teaching methods (such as lecture method, discussion method, project-based learning) based on detailed teaching objectives and student learning analysis results, and designing specific, attractive and interactive teaching activity sequences.

[0139] IV. Course Introduction Design Agent: Specially designed engaging course introduction activities to stimulate students' learning interest and clarify the teacher and student activities and design intentions in the introduction activities.

[0140] V. Concept Explanation Agent: Design clear, easy-to-understand, and multi-faceted explanation strategies and presentation methods for key concepts or difficult points in the teaching content.

[0141] VI. Exercise Generation Intelligent Agent: Based on teaching objectives and content, automatically generate practice questions and test questions of different difficulty levels and in different formats (multiple choice, fill-in-the-blank, short answer, programming, etc.) for classroom practice or homework.

[0142] VII. Teaching Reflection Intelligence: Design teaching effectiveness evaluation schemes (such as classroom observation scales, student work evaluation standards, and questionnaires), and provide a framework and guiding questions for teaching reflection to promote teachers' improvement of teaching.

[0143] This role setting decomposes complex instructional design optimization tasks to specialized intelligent agents, achieving modular division of labor.

[0144] Step S22: Describe the second requirement for different roles of the dedicated intelligent agent.

[0145] Develop specific requirements descriptions (i.e., second requirements descriptions) for each dedicated agent role. Unlike the requirements description for the manager agent (which focuses on "what" and defines the framework), the requirements description for dedicated agents focuses on "how" and includes procedural knowledge to guide the agent in performing specific optimization tasks.

[0146] For example, the requirement description for the "learning objective refinement agent" could be: "Your responsibility is to receive the initial learning objectives and related knowledge points, and, based on Bloom's Taxonomy of Knowledge and the SMART principle, refine them into specific objectives at three levels: knowledge and skills (using verbs such as recognition, description, and application), processes and methods (emphasizing inquiry, cooperation, and expression), and affective attitudes and values ​​(focusing on interest, habits, and a sense of responsibility). Ensure that the objectives are clearly stated and measurable, and analyze the teaching focus and difficulties of each knowledge point."

[0147] Write similar requirements descriptions for each specialized intelligent agent, specifying its input, processing logic, output requirements, and the educational principles or methods to be followed.

[0148] Step S23: Generate the corresponding prompt task framework based on the second requirement description.

[0149] Similar to step S13 above, the second requirement description for each dedicated agent is input into the agent creation platform to generate its own task prompting framework. These frameworks also contain modules such as @Priming, @Persona, @Audience, @ContextControl, and @Instruction, but the instructions in @Instruction will be more specific, focusing on specific optimization tasks.

[0150] Step S24: Hint that the task framework is insufficient for the dedicated intelligent agent to perform fine-grained processing, and build the knowledge base required for the dedicated intelligent agent.

[0151] To enable specialized intelligent agents to optimize based on professional educational knowledge, rather than simply relying on the general knowledge of large language models, it is necessary to build or introduce structured knowledge bases (i.e., instructional design-related knowledge bases). These knowledge bases contain educational knowledge and resources specific to certain domains.

[0152] I. Knowledge Base Construction: By collecting and organizing authoritative textbooks, curriculum standards, educational research literature, excellent teaching cases, question bank resources, etc., the following types of knowledge bases will be constructed:

[0153] The teaching objectives knowledge base includes Bloom's hierarchy of cognitive verbs for common knowledge points in various subjects, SMART principle templates, and examples of three-dimensional objective statements.

[0154] Teaching resource knowledge base: contains links and descriptions of various teaching media (images, videos, animations, simulation experiments), online tools, recommended readings, and other resources.

[0155] Teaching Methods Knowledge Base: Includes definitions, applicable scenarios, implementation steps, and advantages and disadvantages analysis of various teaching methods (such as lecturing, discussion, PBL, flipped classroom).

[0156] Assessment Methods Knowledge Base: Includes methods, tools (scales, scoring criteria) and cases for formative assessment (such as questioning, classroom quizzes, and work analysis) and summative assessment (such as unit tests and final exams).

[0157] Active Learning Technology Knowledge Base: Includes teaching strategies and technologies that promote active student participation (such as mind mapping, concept mapping, group collaboration, and debate).

[0158] Interdisciplinary and General Competency Knowledge Base: Includes interdisciplinary thematic case studies, project-based learning design templates, and methods and activities for cultivating students' critical thinking, innovation, and communication and collaboration skills.

[0159] II. Knowledge Base Format: Knowledge bases can be stored as structured databases, vector databases (for semantic retrieval), or simple text files / tables.

[0160] Step S25: After constructing the knowledge base, upload different knowledge bases in the task prompts of the dedicated intelligent agent.

[0161] On the agent creation platform, the constructed knowledge base is associated with the corresponding dedicated agent. This is typically achieved through the platform's knowledge base upload function. The task prompting framework for each dedicated agent explicitly specifies which knowledge bases it needs to query and utilize when performing a task. For example, the task prompt for a "teaching objective refinement agent" associates the teaching objective knowledge base; the task prompt for a "teaching process generation agent" associates the teaching method knowledge base and the teaching resource knowledge base. The platform will then invoke the relevant knowledge base content based on the prompts when the agent is running.

[0162] Step S26: Create different specialized intelligent agents based on multiple task prompts containing a knowledge base.

[0163] The task hint framework, which includes knowledge base references, and corresponding parameters (which can be adjusted according to task characteristics; for example, a concept explanation agent may require a higher creativity setting and a slightly higher temperature) are published on the agent creation platform, completing the creation of each specialized agent. Now, each specialized agent has the ability to perform its specific optimization task and can call upon the professional knowledge base to improve optimization quality.

[0164] In one exemplary embodiment, the managerial intelligent agent and the dedicated intelligent agent can be integrated to form a collaborative team. Through human-computer interaction and inter-agent calls, a complete instructional design process, from generating the instructional syllabus to optimizing its details, is completed. This embodiment includes the following steps:

[0165] Step S31: Integrate the dedicated intelligent agent into the manager intelligent agent to form an intelligent agent team.

[0166] In the agent team, the manager's task prompts are added to the task allocation module. That is, user input is parsed through a predefined keyword rule table to realize the division of functions between the manager agent and the special agent.

[0167] I. Agent Integration: Utilizing the plugin mechanism, API calls, or internal routing capabilities provided by the low-code platform, at least one pre-created dedicated agent is integrated into the runtime environment of the manager agent. When using the system, the user appears to interact only with the manager agent, but the manager agent can internally invoke the corresponding dedicated agents as needed. This integration method can be static (all connections are pre-configured) or dynamic (loaded on demand based on user needs or task context).

[0168] II. Task Assignment Mechanism: Design a task assignment logic within the prompting framework of the manager agent (especially the @Instruction module). This logic is based on keyword recognition and rule matching.

[0169] Define a keyword rule table: Create a rule table that maps possible user inputs or internal states to corresponding actions (executed by the administrator or invoked by a dedicated intelligent agent). Example of a rule table:

[0170] Input containing "Generate", "Design", "Create" + knowledge points -> triggers the manager agent to generate outline mode.

[0171] Input containing "optimize", "modify", "refine", "improve" + teaching element name (e.g., "teaching objectives", "classroom activities", "exercises") -> triggers the administrator agent to call the corresponding dedicated agent.

[0172] If the score of "Teaching Process" in the internal evaluation results is less than the threshold, the administrator agent will call the "Teaching Process Generation Agent".

[0173] Parsing and Invocation: After receiving user input or completing the initial generation, the manager agent first parses the instructions or evaluates the results through this rule table to determine whether it should handle the next step itself or invoke a dedicated agent.

[0174] Step S32: After realizing the division of functions between the managerial intelligent agent and the dedicated intelligent agent, the instructional design is generated and optimized through the intelligent agent team until all teaching elements meet the teaching requirements.

[0175] Specifically, the teaching knowledge points and student background information are input into the administrator's intelligent agent to generate a teaching design outline. After the teacher reviews the teaching design outline, the administrator's intelligent agent receives feedback information input by the teacher. The administrator's intelligent agent calls the corresponding dedicated intelligent agent to optimize the teaching elements based on the feedback information. It is then determined whether the optimized teaching elements meet the teaching requirements. If not, the optimization process is repeated until each teaching element meets the teaching requirements.

[0176] For example, this is a closed-loop process of iterative optimization:

[0177] I. Initial Generation: The user (teacher) inputs teaching requirements into the administrator agent (e.g., the knowledge point "buoyancy concept in junior high school physics", and the student's background "eighth grade, average physics foundation"). The administrator agent generates an initial teaching design syllabus based on the provided framework.

[0178] II. Manual Review and Feedback: Teachers review the generated lesson plans and identify areas for improvement. For example, teachers may feel that the learning objectives are not specific enough or the classroom activities are not engaging enough. Teachers input specific feedback (such as "Please optimize the learning objectives to make them more measurable" or "Please add a demonstration experiment to explain the concept of buoyancy") into the administrator's AI agent.

[0179] III. Agent Collaboration Optimization: After receiving feedback information, the manager agent analyzes it using the function allocation mechanism established in step S31.

[0180] If the feedback is for optimizing a specific element (such as "optimizing teaching objectives"), the administrator agent recognizes the keywords "optimization" and "teaching objectives" and then invokes the "teaching objective refinement agent". The invoked dedicated agent receives relevant contextual information (such as the original objectives, teacher feedback, and knowledge points) as well as its own task prompts and knowledge base, executes the optimization task, and generates the optimized teaching objectives.

[0181] The managerial agent integrates the optimized results back into the instructional design syllabus.

[0182] IV. Iteration and Judgment: Teachers review the optimized instructional design again.

[0183] Judgment criteria: Teachers judge whether the optimized elements are satisfactory based on their professional experience and teaching requirements (such as whether they meet curriculum standards, are feasible, and can achieve the expected teaching results). The system can also be configured with automated checking rules (such as SMART principle verification and activity type coverage checks).

[0184] Iterative optimization: If the teacher believes further improvement is needed, they can continue to provide feedback, and the administrator agent will again invoke the corresponding dedicated agent for optimization. This "feedback-invocation-optimization-review" cycle can be repeated multiple times.

[0185] Termination condition: When the teacher believes that all teaching elements have met the teaching requirements (or the system's internal checks have passed), the teacher stops providing negative feedback or explicitly expresses satisfaction, and the optimization process ends.

[0186] V. Final Integration: The administrator agent integrates all optimized and confirmed satisfactory teaching elements into a complete, coherent, and high-quality instructional design scheme, which is then provided to teachers for use.

[0187] In one exemplary embodiment, such as Figure 3 As shown, this embodiment provides an instructional design generation system. The system can be deployed on a server, personal computer, or cloud platform, and includes components such as a processor, memory, and network interface. At the software level, the system includes the following functional modules:

[0188] A manager intelligent agent module 100 is constructed to acquire instructional design data and extract instructional elements; based on the instructional elements, a first requirement description is performed on the instructional design to construct a prompting framework; the prompting framework is optimized to ensure that the instructional elements meet the instructional requirements; a manager intelligent agent is created based on the optimized prompting framework, and the manager intelligent agent is used to generate an instructional design outline according to the input instructional knowledge points;

[0189] A dedicated intelligent agent module 200 is constructed to design multiple optimization elements according to the teaching design outline, and to set different roles for the teaching design based on the multiple optimization elements; to describe the different role settings in a second requirement to construct different task prompts; to call the knowledge base related to the corresponding task according to the different task prompts; and to create different dedicated intelligent agents based on multiple task prompts including different knowledge bases.

[0190] The multi-agent collaborative design teaching module 300 is used to call at least one of the dedicated agents by the manager agent to optimize specific teaching elements in the teaching design outline and generate a complete teaching design scheme that meets educational standards.

[0191] Through the collaborative work of the above modules, the system realizes a method for generating instructional designs, which can efficiently produce high-quality instructional design schemes.

[0192] The instructional design generation system in this application embodiment can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. For example, mobile electronic devices can be mobile phones, tablets, laptops, PDAs, in-vehicle electronic devices, wearable devices, ultra-mobile personal computers (UMPCs), netbooks, or personal digital assistants (PDAs), etc., while non-mobile electronic devices can be servers, network-attached storage (NAS), personal computers (PCs), etc. This application embodiment does not impose specific limitations.

[0193] One example of the instructional design generation system in this application is a device with an operating system. This operating system can be Android, iOS, or other possible operating systems; this application does not specifically limit the specific operating system used.

[0194] The instructional design generation system provided in this application embodiment can realize all the processes of the above-mentioned instructional design generation method. To avoid repetition, it will not be described again here.

[0195] Optionally, embodiments of this application also provide an electronic device, including a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the various processes of the above-described instructional design generation method embodiment and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0196] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described embodiment of the instructional design generation method and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0197] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0198] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0199] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0200] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for generating instructional designs, characterized in that, The instructional design generation method, applied to a first device on which a first intelligent agent is deployed, includes: In response to the user's input of a requirement description, an initial prompt corresponding to the requirement description is generated; Determine a first number of teaching elements in the requirements description, and determine a second number of teaching elements in the initial prompt; The second quantity is scored based on the first quantity to obtain a first sub-score; If the first sub-score is greater than or equal to the first score threshold, then the generation order of each teaching element in the initial prompt and the corresponding prompt text are determined. The generation order is scored to obtain a second sub-score; The content consistency of the prompt texts corresponding to each adjacent teaching element is scored separately to obtain the third sub-score; If the sum of the second sub-score and the third sub-score is greater than or equal to the second score threshold, then the initial prompt is optimized to obtain the target prompt; the target prompt carries at least a teaching element. Based on the target prompts, the large language model is invoked to generate the teaching design outlines corresponding to the teaching elements; From the second agent deployed on the second device, the third agent corresponding to each teaching element is invoked to generate the optimized text corresponding to the initial text of each teaching element in the teaching design outline; each second agent can search in the teaching design-related knowledge base corresponding to the teaching element. The aforementioned instructional design outline and all optimized texts are integrated to obtain the target instructional design.

2. The instructional design generation method according to claim 1, characterized in that, The step of scoring the second quantity based on the first quantity to obtain a first sub-score further includes: If the first sub-score is less than the first score threshold, then return to the initial prompt step corresponding to generating the requirement description.

3. The instructional design generation method according to claim 1, characterized in that, The process of scoring the content consistency of the prompt texts corresponding to each adjacent teaching element to obtain a third sub-score, followed by: If the sum of the second sub-score and the third sub-score is less than the second score threshold, then return to the initial prompt step corresponding to generating the requirement description.

4. The instructional design generation method according to any one of claims 1 to 3, characterized in that, The optimization of the initial prompt to obtain the target prompt includes: Determine the user's requirement for secondary input; Based on the requirements described by the secondary input, the large language model is invoked to optimize the initial prompt, thereby obtaining the target prompt.

5. A method for generating instructional designs, characterized in that, The instructional design generation method, applied to a second device with a second intelligent agent deployed thereon, and for any third intelligent agent among the second intelligent agents, includes: In response to a call instruction sent by a first intelligent agent deployed on a first device, the initial text corresponding to the target teaching element in the instructional design syllabus carried in the call instruction is determined; the target teaching element corresponds to any of the third intelligent agents. The initial text is optimized to generate optimized text; The optimized text is sent to the first intelligent agent; the optimized text is used to integrate with the instructional design outline to obtain the target instructional design.

6. The instructional design generation method according to claim 5, characterized in that, The process of optimizing the initial text to obtain optimized text includes: Based on the target teaching elements and the initial text, a search is performed in the instructional design-related knowledge base to obtain the recall text; The initial text is optimized based on the recalled text to obtain the intermediate text; If the intermediate text meets the optimization requirements, the intermediate text is determined to be the optimized text.

7. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein when the program or instructions are executed by the processor, they implement the steps of the instructional design generation method as described in any one of claims 1-4 or 5-6.

8. A readable storage medium, characterized in that, The program or instructions are stored on the readable storage medium, and when the program or instructions are executed by a processor, they implement the steps of the instructional design generation method as described in any one of claims 1-6.

Images